Data-Driven Retrospectives: Go Beyond Opinion
retrospective data metrics
“What do you think of the sprint?” — opinions vary depending on mood, personality, and selective memory. Data tells an objective story that complements and challenges perceptions.
Why Data in Retrospectives
- Selective memory — we remember the most recent bug, not the 90% that went fine
- Recency bias — what happened yesterday seems more important than what happened 2 weeks ago
- Personality — extroverts dominate, introverts stay quiet
- Endless discussions — “I think that” vs “the data show that”
Essential Data for Retrospectives
1. Sprint Deliveries
| Metric | Previous Sprint | Current Sprint | Trend |
|---|---|---|---|
| Planned points | 25 | 24 | → |
| Delivered points | 22 | 24 | ↑ |
| Carry-over | 3 points | 0 | ↓ |
2. Quality
| Metric | Previous Sprint | Current Sprint | Trend |
|---|---|---|---|
| Production bugs | 2 | 4 | ↑ |
| Critical bugs | 0 | 1 | ↑ |
| PRs without review | 1 | 0 | ↓ |
3. Velocity and Flow
- Velocity of last 5 sprints: 20, 22, 18, 25, 24
- Trend: stable, slightly upward
- Average lead time: 4.2 days (previously: 5.1 days)
4. Estimate vs. Actual
| Item | Estimated | Actual | Difference |
|---|---|---|---|
| PIX Checkout | 8 | 13 | +63% |
| User profile | 3 | 3 | 0% |
| Advanced search | 5 | 8 | +60% |
| Notifications | 3 | 2 | -33% |
5. Team Health
- Absences in sprint: 2 days
- Overtime: 0 (ideal)
- Overall feeling: 3.8/5 (quick survey)
Tools for Collecting Data
- Jira/Velocity Chart — velocity and burndown
- GitHub/GitLab — PRs merged, lead time, review time
- SonarQube — code quality
- Dev in Poker — estimate history
- Quick Survey — 3 questions at the end of the sprint
- Datadog/Sentry — production bugs and incidents
Data-Driven Retrospective Format
- Present the data (5 min) — Scrum Master shows the numbers
- Reactions (5 min) — “What surprises you about the data?”
- Insights (10 min) — what caused each variation?
- Actions (5 min) — maximum 2 actions for the next sprint
- Check-out (2 min) — how do you feel about the sprint?
Example of a Data-Driven Insight
Without data: “I feel like we’re delivering less.” With data: “We delivered 24 points vs 22 from the previous sprint, but we had 4 production bugs vs 2. Quality dropped even though delivery was stable. Why?”
Action: “Investigate why bugs doubled and create a pre-release QA checklist.”
Retrospective Health Metrics
The retrospective itself can be measured:
- Implemented actions from the previous sprint (how many out of how many?)
- Participation — did everyone contribute?
- Time — did it stay within the timebox?
- Follow-up — are actions visible and tracked?
Conclusion
Data transforms retrospectives from a “complaint session” into a “fact-driven diagnosis.” Opinions still matter — data provides the context so that those opinions are informed and actionable.